Selection from the Medically Needy Program

I begin with qualitative evidence on risk selection in insurance coverage and future Medicaid utilization. Figure 1.3 presents private insurance coverage (top) and future Medicaid utilization (bottom) rates, adjusted for year and age, as a function of health risk for individuals in my sample prior to 2006 and living in states with no policy changes in 2002 or 2004. The left axis is the share of the population with either insurance coverage or who use Medicaid in the future, while the right axis is the rate in states with Medically Needy programs minus the rate in states without such programs. The top graph demonstrates that private insurance coverage is less common for individuals at higher risk in states with Medically Needy programs, with coverage decreasing from 75% to 65% at the highest level of risk. The relationship between private insurance and risk for individuals living in states without a Medically Needy program is noisy, with the decrease in coverage in the middle of the risk distribution reflecting a ten percentage point decrease in Medigap

Figure 1.3: Insurance coverage and health

Notes: Graphs are locally weighted average residuals from regression of private insurance coverage (top) or future Medicaid (bottom) on year and age dummies. Horizontal axis is percentile in the health risk distribution. Line labeled difference is the Medically Needy minus not Medically Needy locally weighted average residual.

coverage, while Medicare Advantage coverage rates are steady across the risk distribution (not shown). The bottom graph demonstrates that there is little difference in future Medicaid utilization at low levels of health risk. But in the upper half of the risk distribution, future Medicaid utilization is increasing more rapidly for individuals living in states with Medically Needy programs.

Figure 1.4 illustrates trends over time in private insurance coverage and future Medicaid

uti-Figure 1.4: Time Trends in Private Insurance Coverage and Medicaid Utilization

Note: Point estimates based on full sample, vertical lines are 95% confidence intervals.

lization for individuals living in states with and without Medically Needy programs. Prior to the introduction of Medicare Part D in 2006, the trends in private insurance coverage are different for individuals in each set of states. However, this is due solely to differences in Medicare Advan-tage enrollment—enrollment in Medigap is virtually identical in states with and without Medically Needy programs in all years. Tends in future Medicaid utilization are similar prior to 2006.

Figure 1.3 provides qualitative evidence in favor of selection induced by Medically Needy pro-grams. I begin my quantitative tests by estimating the regression equation in (1.1); the results are presented in table 1.2. In the odd-numbered columns, I include controls for how supplemental private insurance plans are priced—gender, smoking status, age, and state²³—which permits me to interpret the coefficients on insurance status as a positive correlation test for asymmetric in-formation (Chiappori et al., 2006; Chiappori and Salani, 2000). In the even-numbered columns, I include fixed effects for each demographic group used in constructing the simulated instrument and marital status, in addition to the covariates in the first set of regressions. The first two columns show that private insurance is advantageously selected, consistent with previous work. But, the next two columns demonstrate that private insurance is advantageously selected only in states with Medically Needy programs. The introduction of Medicare Part D, which lowered out-of-pocket spending per unit of health risk, lead to an increase in risk among the privately insured in states with Medically Needy programs, although this was primarily due to a reduction in risk in states without Medically Needy programs. These results are consistent with the Medically Needy pro-gram inducing advantageous selection in the private insurance market.

The next four columns test if the deductible affects risk selection. I instrument for the de-ductible using the simulated dede-ductible (see Appendix Table 1.B for the first stage regressions associated with estimating (1.2) in table 1.3).²⁴ Overall the coefficients on the deductible and its interaction with insurance coverage do not provide evidence that the financial cost of using the Medically Needy program is an important contributor to selection. The negative coefficient on the deductible in column (5) provides an indication that individuals who are wealthier, since wealth is an important determinant of the deductible, are also healthier, but this effect does not persist after I include a more exhaustive set of controls for demographic characteristics.

23Medigap plans are priced at the zip code level, but Maestas et al. (2009) find very little within-state variation in premiums, so state fixed effects should be sufficient to control for the prices of contracts that insurers offer to individuals (Fang et al., 2008).

24The two-stage least squares coefficient estimates on the deductible differ in specifications that do not include demographic fixed effects from specifications that do include demographic fixed effects because of a mechanical correlation between those fixed effects and the simulated deductible. In the absence of demographic fixed effects, the local average treatment effect (Angrist et al., 1996) reflects two sources of variation—state changes in eligibility rules and differences between demographic groups—but including demographic fixed effects ensures that the treatment effect is only due to state changes in eligibility rules. That does not mean that the difference in these two treatment effects is unimportant, rather these differences demonstrate that individuals who belong to demographic groups that have systematically higher deductibles, due to larger wealth holdings, are also in better health.

Table1.2:RiskSelectionfromStateMedicallyNeedyPrograms (1)(2)(3)(4)(5)(6)(7)(8) Insured-0.063+-0.065*0.0370.0180.0340.0140.0340.014 (0.034)(0.032)(0.051)(0.049)(0.050)(0.050)(0.050)(0.050) PartD×Insured-0.005-0.004-0.151+-0.141-0.150+-0.140-0.150+-0.138 (0.048)(0.047)(0.082)(0.088)(0.080)(0.086)(0.080)(0.086) MedicallyNeedy×Insured-0.129*-0.108+-0.096-0.101+0.019-0.014 (0.061)(0.058)(0.061)(0.060)(0.111)(0.117) MN×PartD×Insured0.193*0.180+0.198*0.189+0.1870.148 (0.095)(0.102)(0.096)(0.104)(0.127)(0.138) Deductible-0.023**-0.005-0.0040.009 (0.008)(0.011)(0.016)(0.017) Deductible×PartD-0.004-0.006-0.007-0.013 (0.010)(0.011)(0.015)(0.017) Deductible×Insured-0.029-0.022 (0.025)(0.027) Deductible×PartD×Insured0.0010.009 (0.021)(0.025) R-squared0.040.090.040.090.030.010.030.01 N72397239723972397238723872387238 Fonexcludedinstruments143.3163.137.459.0 +p<0.1,*p<0.05,**p<0.01 Notes:Dependentvariableisthestandardizedriskscore,AnyPrivateisanindicatorforhavingsomeformofsupplementaryinsuranceinadditionto Medicare,PartDisanindicatorfor2006and2008.Pricingcontrolsareacubicpolynomialinage,gender,andsmokingstatus.Demographicfixed effectsareindictorsforeachuniquecombinationofgender,race,ethnicity,andeducation,seetextforadditionaldetails.Allmodelsincludestateand yearfixedeffects;modelswithdemographicfixedeffectsalsoincludemaritalstatus.Instrumentsforthedeductibleanditsinteractionsarethe simulatedaveragedeductibleanditsinteractions.Standarderrorsclusteredonstateinparentheses.

I next turn to the effect of the Medically Needy program on the decision to purchase private coverage. Table 1.3 presents results based on the regression in (1.2). The first column demonstrates that the Medically Needy program does not reduce private insurance coverage. Without a change in private insurance coverage, either higher or lower, the only way to reduce average risk among the privately insured is if high risk individuals are less likely to purchase insurance, while low risk individuals are more likely to purchase insurance in states with Medically Needy programs. Col-umn 2 shows that this is the case—the probability of purchasing private insurance is 4.2 percentage points lower for a one unit increase in risk. Assuming that the demand for insurance for low risk individuals is decreasing in price, the implication is that the Medically Needy program reduces premiums .²⁵

The remaining four columns incorporate the Medically Needy deductible, allowing the effect of the Medically Needy to vary among individuals for whom the program is more or less generous.

The OLS results (columns 3 and 5) indicate that individuals with higher deductibles are more likely to have private insurance, although once I instrument for the deductible, using the simulated deductible, the effect of the deductible is no longer statistically significant. The deductible becomes less relevant after the introduction of Medicare Part D because, by lowering the price of risk, Medicare Part D makes it more difficult to enter the Medically Needy program. The result is that lower deductible individuals should be more likely, at the margin, to purchase private insurance after Medicare Part D.

The previous discussion pools results for two distinct types of supplemental insurance. In table 1.4 I present results from estimating (1.1) and (1.2) separately for Medigap and Medicare Advan-tage coverage. Panel A, which presents results for risk selection from (1.1), suggests that both types of insurance are advantageously selected in states with Medically Needy programs, but it is only Medigap plans for which risk is higher in states with higher deductibles. Rather, individu-als with higher deductibles and Medicare Advantage coverage are healthier than individuindividu-als with lower deductibles. Panel B, corresponding to enrollment based on (1.2), indicates that the Medi-cally Needy program has offsetting effects on enrollment into Medigap and Medicare Advantage, although neither are statistically significant, and that the bulk of the response to the incentives of

25If the premium were unaffected by the Medically Needy program there would be no effect on enrollment by low-risk individuals and the coefficient on Medically Needy in column (1) would be negative.

Table1.3:EffectoftheMedicallyNeedyProgramonDemandforPrivateInsurance (1)(2)(3)(4)(5)(6) MedicallyNeedy0.0040.024-0.023-0.034-0.003-0.014 (0.027)(0.031)(0.031)(0.048)(0.034)(0.052) MN×PartD-0.016-0.044-0.0170.045-0.046+0.019 (0.025)(0.028)(0.025)(0.037)(0.028)(0.039) MN×Risk-0.046**-0.043**-0.042* (0.017)(0.017)(0.017) MN×PartD×Risk0.057**0.056*0.050* (0.022)(0.022)(0.021) Deductible0.006**0.0080.006**0.007 (0.002)(0.008)(0.002)(0.008) Deductible×PartD0.001-0.014*0.001-0.014* (0.002)(0.006)(0.002)(0.006) R-squared0.130.130.130.010.140.01 N723972397239723872397238 Fonexcludedinstruments152.9141.3 +p<0.1,*p<0.05,**p<0.01 Notes:DependentvariableisdummyforhavingsomeformofsupplementaryinsuranceinadditiontoMedicare,PartDisanindicatorfor2006and 2008,Riskisthestandardizedhealthriskscore,DeductibleistheamounttheindividualmustspendtobeeligibleforcoverageundertheMedically Needyprogram.Allmodelsincludeyear,state,anddemographicgroupfixedeffects,maritalstatus,acubicpolynomialinage,andanindicatorfor smoking.InstrumentsforthedeductibleandthedeductibleinteractedwithPartDarethesimulatedaveragedeductibleandthesimulatedaverage deductibleinteractedwithPartD.Standarderrorsclusteredonstateinparentheses.

Table 1.4: Effect of Medically Needy Programs on Medigap and Medicare Advantage

Medigap Medicare Advantage

(1) (2) (3) (4)

Panel A: Risk

Medically Needy × Insured -0.035 -0.160 -0.076 0.096

(0.088) (0.111) (0.088) (0.127)

MN × Part D × Insured 0.059 0.168 0.131 0.001

(0.076) (0.156) (0.136) (0.192)

Deductible × Insured 0.027+ -0.039+

(0.014) (0.021)

Deductible × Part D × Insured -0.021 0.028

(0.030) (0.031)

Medically Needy × Part D × Risk 0.027 0.023

(0.020) (0.023)

Notes: Dependent variable in Panel A is the standardized risk score, and in Panel B is an indicator for having the specific form of insurance. Insured is an indicator for having the specified form of supplementary insurance in addition to Medicare, Part D is an indicator for 2006 and 2008. All models include state, year, and demographic fixed effects, smoking status, marital status, and a cubic polynomial in age. Instruments for the deductible and its

interactions are the simulated average deductible and its interactions. Standard errors clustered on state in parentheses.

the Medically Needy program comes from individuals deciding if they will enroll in a Medicare Advantage plan, or not.

If individuals are responding to the incentives provided by the Medically Needy program, one would expect that individuals in states with a Medically Needy program are more likely to use Medicaid in the future and that future Medicaid users are in worse health now since eligibility through the Medically Needy program is (partially) dependent on health. Table 1.5 estimates equations (1.1) and (1.2) for future Medicaid utilization. The first three columns demonstrate that individuals who use Medicaid in the future are systematically less healthy in Medically Needy states than similar individuals in states without a Medically Needy program—although this trend reverses with the implementation of Medicare Part D, because future Medicaid users in states without Medically Needy programs become considerably less healthy after Part D.

Individuals in states with Medically Needy programs are also more likely to use Medicaid in the future, relative to individuals living in states without such programs. The 2.6 percentage point increase in utilization, relative to a mean of 6.2%, is large, but it is not implausible given that more than 30% of the sample does not have any supplemental insurance and an individual is considered to have used Medicaid if she has used Medicaid for even one month over the previous two years.²⁶ Assuming that the additional users of Medicaid in the future gain Medicaid coverage through the Medically Needy program, the increase in utilization, combined with the increase in average risk, indicates that users of the Medically Needy program are in substantially worse health now than individuals who do not go on to use the Medically Needy program. The increase in average risk and the increase in demand given an increase in risk imply that either enrollment is growing and the growth in enrollment is among individuals who are sicker than the inframarginal enrollee, or enrollment is falling and the decline is due to individuals who are healthier than the inframarginal enrollee deciding not to use Medicaid; the increase in future Medicaid utilization indicates that the marginal users are sicker than the inframarginal enrollees. Lastly, when I control for the Medically Needy deductible, the effect of risk on insurance coverage is sharply attenuated, reflecting a negative correlation between the deductible and health risk. Given that the deductible is a function of assets, the implication is that individuals in states with Medically Needy programs are not engaging in precautionary savings.

I test the final implication from Table 1.5—that means-tested public insurance programs distort

26These effects can be scaled to the monthly hazard rate that an individual uses Medicaid, which is 0.27% at the mean, and a 2.6 percentage point increase in future Medicaid utilization increases the hazard rate to 0.38%.

Table1.5:FutureMedicaidCoverageandtheMedicallyNeedyProgram HealthRiskMedicaidUtilization (1)(2)(3)(4)(5)(6) MedicallyNeedy-0.071-0.081-0.0220.026*0.0230.035 (0.081)(0.083)(0.120)(0.013)(0.017)(0.027) MN×PartD0.0110.0540.044-0.0090.0170.042 (0.057)(0.063)(0.078)(0.015)(0.012)(0.027) MN×FutureMedicaid0.283*0.249+ (0.138)(0.143) MN×PartD×FutureMedicaid-0.688**-0.688** (0.216)(0.212) Deductible-0.013-0.002 (0.014)(0.004) Deductible×PartD0.001-0.006 (0.013)(0.005) MN×Risk0.023*0.022* (0.011)(0.011) MN×PartD×Risk-0.054*-0.057** (0.022)(0.021) R-squared0.110.110.020.110.120.02 N438843884386438843884386 Fonexcludedinstruments86.484.3 +p<0.1,*p<0.05,**p<0.01 Notes:Dependentvariableincolumns1-3isthestandardizedriskscore,columns4-6isanindicatorforusingMedicaidanytimeinthenexttwoyears. FutureMedicaidisanindicatorforusingMedicaidanytimeinthenexttwoyears,PartDisanindicatorfor2006(2008isexcludedtoconstructFuture Medicaid),MNisshort-handforMedicallyNeedy.Allmodelsincludeyear,state,anddemographicgroupfixedeffects,maritalstatus,acubic polynomialinage,anindicatorforsmoking,andmaineffectsforallinteractions.Instrumentsforthedeductibleanditsinteractionsarethesimulated averagedeductibleanditsinteractions.Standarderrorsclusteredonstateinparentheses.

savings behavior (Hubbard et al., 1995; Kotlikoff, 1986; Levin, 1995)—using (1.3) and present the estimates in levels for total wealth, countable and exempt assets, and equity in the primary residence, and countable assets and equity in the primary residence as a share of total wealth in Table 1.6. The results demonstrate that the Medically Needy program reduces countable assets and induces a shift in assets towards those that are exempt from the deductible calculation. The shift towards home equity is not complete, i.e. the growth in exempt assets is not only due to an increase in home equity. This reflects the fact that one can only reduce mortgage debt, which is the easiest way to increase home equity, up to the outstanding balance of the mortgage. Finally, the effect of an increase in the simulated deductible indicates that asset shifting is most common for individuals in demographic groups that should have low deductibles—i.e. those who are most likely to be able to use the Medically Needy program.